A Hybrid Approach to Software Defect Prediction with Stacking Classifier

Abstract

The prediction of software defects is essential for improving software quality and reducing the cost of testing by determining problematic modules for focused testing. This paper Examines Sophisticated Methodologies, Including the “Synthetic Minority Over-Sampling Technique (Smote) and Particle Swarm Optimization (PSO), Tackle Difficulties Related to Class Impacts and Feature Selection, Utilizing Benchmark Datasets Such As CM1, JM1, JM1 MC2, MW1, PC1, PC3, and PC4”. A stacking classifier that combines a tree of decision -making tree, random forests and lightGBM was used, and works quite well on all data files. The proposed method focuses on the use of a file learning to improve the accuracy of prediction by merging the best parts of several classifiers. Comprehensive evaluation shows that the proposed model is strong and reliable and finds that it is better to find problematic software modules than other models. The aim of this effort is to improve the prediction of defects in the software by offering cost -effective and efficient ways to find defects in time, which will lead to better quality of software and better use of resources during testing.